Modeling stellar convective transport with plumes: II. Transport Properties of Locally and Non-locally driven Convection

arXiv:2511.12824 · astro-ph.SR, physics.flu-dyn · Submitted 2025-11-16 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Astrophysics Radio. Generated commentary on the latest astrophysics papers.

Vera: I'm Vera.

Jocelyn: And I'm Jocelyn, here with Subrahmanyan.

Vera: We're looking at the paper: "Modeling stellar convective transport with plumes".

Jocelyn: Here's the gist of the paper. Modeling stellar convective transport with plumes: II.

Vera: Transport Properties of Locally and Non-locally driven Convection performs three-dimensional hydrodynamic simulations of two idealized regimes of stellar convection: a cooling-driven model (Model C) and an entropy-gradient–driven model (Model S).

Jocelyn: The two regimes exhibit striking contrasts: while Model S develops large, relatively stationary eddies excited at depth, Model C is dominated near the surface by intermittent plume-like downflows that produce broad non-Gaussian velocity distributions and a turbulent energy flux that exceeds Model S by nearly an order of magnitude in the upper convection zone Contrasting Convection Regimes The study investigates two contrasting convection models: the cooling-driven model (Model C) and the entropy-gradient–driven model (Model S) Model S develops large, relatively stationary eddies excited at depth, while Model C is dominated near the surface by intermittent plume-like downflows that produce broad non-Gaussian velocity distributions and a turbulent energy flux that exceeds Model S by nearly an order of magnitude in the upper convection zone The two models are similar to those studied by Cossette & Rast (2016, CR16), except that CR16 included the stably stratified layer beneath the CZ, whereas we restrict the computational domain to the CZ itself In Model C, plume-like downflows are spontaneously generated near the surface, whereas large-scale convective flows from deeper layers are absent Statistical Signatures of Plume Dynamics The statistical properties reveal significant differences between the two regimes For Model C, the probability density of vertical velocity reveals a clear model dependence where the downflow wing is substantially broader than in Model S The joint probability distributions of vertical velocity and vertical turbulent energy flux demonstrate that frequent, high-speed downflows dominate the transport, exhibiting a nonlinear scaling Specifically, the turbulent energy flux is approximately proportional to δuz3 in the downflow regime Failure of Local Gradient-Diffusion Models The conventional gradient–diffusion (GD) hypothesis fails to describe Model C accurately The GD model underestimates transport in the upper CZ because the mean entropy gradient is small, even though vigorous plumes carry substantial energy This discrepancy clearly demonstrates that the cooling-driven convection cannot be described by a simple gradient-diffusion closure Development of a Modified Closure To address this failure, the authors introduce a modified Gradient–Diffusion approach incorporating plume dynamics via the Time–Space Double Averaging (TSDA) method They propose a modified GD closure where the turbulent diffusivity is corrected by a plume-mediated term proportional to ⟨(u˜·∇)u′2⟩h This correction term is negative in the upper CZ, implying an inward transport of incoherent kinetic energy that enhances the effective turbulent energy transport coefficient Astrophysical Implications The results suggest that surface cooling generates coherent, intermittent plumes whose dynamics feed on—and redistribute—the energy of incoherent fluctuations The predicted non-Gaussian statistics, including broad downflow tails and cubic flux–velocity scaling, provide concrete targets for observational tests using helioseismic inversions and high-cadence flow diagnostics This modified closure offers a physically motivated extension applicable to non-equilibrium stellar convection zones Limitations and Future Directions Two main limitations are noted: the magnitude of the non-equilibrium correction depends on the time-averaging window T used to define coherence Furthermore, model constants like α cannot yet be predicted from fundamental equations and must be calibrated against data Future work will survey parameter space and include rotation/MHD to clarify applicability beyond the solar CZ The main findings include: (i) Contrasting convection morphologies: Model C is dominated in the upper CZ by intermittent, plume-like downflows generated by surface cooling, whereas Model S exhibits larger, more stationary eddies excited at depth (ii) Marked non-Gaussianity in Model C: Probability densities of the vertical velocity develop broad downflow wings in Model C, and the joint probability distributions of δuz and δeiδuz demonstrate that frequent, high-speed downflows dominate the transport (iii) Enhanced turbulent energy flux near the surface: The vertical turbulent energy flux in Model C exceeds that of Model S by up to an order of magnitude in the upper CZ (iv) Breakdown of local GD closures: While the GD approximation reproduces Model S fairly well, it underestimates the flux in Model C by nearly an order of magnitude in the upper CZ (v) Role of coherent motions and TSDA: Using TSDA, the coherent component u˜z peaks where the flux is largest in Model C, demonstrating that plume coherence mediates transport (vi) A modified non-equilibrium closure: A modified GD closure successfully reproduces both the amplitude and depth-dependence of the flux in Model C when parameters are appropriately specified The TSDA analysis together with the success of the modified closure points to a consistent physical picture: surface cooling generates coherent, intermittent plumes whose dynamics feed on—and redistribute—the energy of incoherent fluctuations The resulting non-equilibrium transport cannot be captured by a purely local, gradient–based flux; instead, it depends on plume coherence and intermittency By embedding plume dynamics into the closure, the modified GD framework successfully bridges the gap between numerical simulations and theoretical modeling Final Caveats The present closure still depends on the choice of T and on a calibrated α Developing strategies to tie these parameters to observable plume statistics will be an important direction for future work Acknowledgements Numerical computations were carried out on Cray XC50 at Center for Computational Astrophysics (CfCA), National Astronomical Observatory of Japan (NAOJ) This work was supported by the Research Institute of Stellar Explosive Phenomena at Fukuoka University & also from the University grant No.GR2302, and also by JSPS KAKENHI Grant Number (JP17H06364, JP18H01212, JP18K03700, JP21K03612, JP21H01088, JP23K22494, JP23K25895, JP23K03400, JP24K00631 and JP25K07374) This research was also supported by MEXT as “Program for Promoting researches on the Supercomputer Fugaku” (Structure and Evolution of the Universe Unraveled by Fusion of Simulation and AI; Grant Number JPMXP1020230406) and JICFuS The National Institutes of Natural Sciences (NINS) program for crossdisciplinary study on Turbulence, Transport, and Heating Dynamics in Laboratory and Solar/Astrophysical Plasmas: “SoLaBo-X” supports this work Data Availability The data underlying this article will be shared on reasonable request to the corresponding author The paper is typeset from a TEX/LATEX file prepared by the author The references list includes authors such as Masada Y., Sano T., and Yokoi N.

Vera: The work is supported by grants from Fukuoka University and the National Astronomical Observatory of Japan The paper was accepted under XXX and received in original form ZZZ This paper has been typeset from a TEX/LATEX file prepared by the author The authors are Youhei Masada, Tomoya Takiwaki, and Nobumitsu Yokoi The paper was accepted under XXX.

Title and authors: Vera: So, this paper is titled "Modeling stellar convective transport with plumes: II. Transport Properties of Locally and Non-locally driven Convection". It's about looking at how heat moves in stars using three dimensional hydrodynamic simulations of two different convection models, Model C and Model S.

Jocelyn: Right, so the title tells us we're looking at transport properties in these models that are driven by local versus non-local mechanisms. It sets up a comparison between two distinct ways convection can happen inside a star.

Subrahmanyan: What does this actually mean for the big picture? We're comparing a cooling-driven model, Model C, against an entropy gradient driven model, Model S. These are idealized scenarios to see how different physics leads to very different transport behaviors in the convection zone.

Vera: Exactly. The key contrast is that Model S develops these large, relatively stationary eddies deep inside the star. But Model C is dominated near the surface by intermittent plume-like downflows that create broad velocity distributions and a turbulent energy flux that's almost ten times bigger than in Model S up in the upper convection zone.

Jocelyn: That difference in flux is striking; it suggests that surface cooling generates something very different from those deep eddies we see in Model S. It’s not just a small tweak to the physics, it’s a big change in how energy moves there.

Subrahmanyan: And what the paper highlights is that these plume-like downflows in Model C produce non Gaussian velocity distributions. That means we can't just use simple average values to describe the flow; we need something more complex to capture what's actually happening with those plumes.

Vera: That points directly to the statistical signatures they found. The probability density of vertical velocity in Model C shows a clear dependence on which model you are looking at, specifically that the downflow wing is substantially broader than in Model S.

Jocelyn: And when you look at the joint probability distributions of vertical velocity and vertical turbulent energy flux, it’s clear that frequent high speed downflows are what really dominate the transport mechanism.

Title and authors: Subrahmanyan: They even found a nonlinear scaling relationship there; specifically, they show that the turbulent energy flux is approximately proportional to delta u z cubed in the downflow regime. That cubic scaling is a very specific piece of information about how these plumes are moving and carrying energy.

Vera: Now, this is where things get interesting because they point out a big problem with older theories. The conventional gradient diffusion, or GD, hypothesis just doesn't describe Model C accurately at all.

Jocelyn: The reason it fails is that the mean entropy gradient is actually quite small in the upper convection zone, even though these vigorous plumes are carrying a lot of energy. So the GD model underestimates transport by nearly an order of magnitude there.

Subrahmanyan: This discrepancy shows that you can't just use a simple gradient diffusion closure to describe cooling-driven convection; you need to account for something more dynamic happening at the surface.

Vera: To fix this, they introduce a modified Gradient Diffusion approach that incorporates plume dynamics using the Time Space Double Averaging method. They propose correcting the turbulent diffusivity by adding a term proportional to the average of u tilde dot grad u prime squared in Model C.

Jocelyn: That correction term is negative in the upper convection zone, which implies an inward transport of incoherent kinetic energy that actually enhances the effective turbulent energy transport coefficient. It’s a clever way to fix that underestimation they found earlier.

Subrahmanyan: This modification successfully bridges the gap between the simulation results and a theoretical framework, showing how coherent plume dynamics feed on and redistribute the energy from those smaller, incoherent fluctuations.

Vera: So what does this mean for us observing stars? The paper suggests that surface cooling generates these coherent, intermittent plumes whose dynamics actively feed on and redistribute the energy from the more random fluctuations we usually deal with.

Title and authors: Jocelyn: It gives us concrete targets for future observations. They predict non Gaussian statistics, including those broad downflow tails and that cubic flux velocity scaling. That means we can use helioseismic inversions or high cadence flow diagnostics to test these specific signatures.

Subrahmanyan: These predicted statistics give observational astronomers a way to confirm if this picture of surface cooling driving coherent plumes is correct, moving beyond just looking at the average properties of the flow.

Vera: There are two limitations they flag, though. First, the magnitude of that non equilibrium correction depends on the time averaging window T you choose to define coherence. Second, some model constants like alpha can't be predicted from fundamental equations yet and have to be calibrated using data.

Jocelyn: That means we don't have a perfectly fixed answer right now; it still depends on how long you look at the flow and how well we calibrate those initial parameters.

Subrahmanyan: Future work needs to survey more parameter space and include rotation or magnetic fields to see if this physics holds up beyond just the solar convection zone. That’s where things get more complex.

Vera: So, to wrap up, the paper "Modeling stellar convective transport with plumes: II. Transport Properties of Locally and Non-locally driven Convection" shows that surface cooling generates coherent, intermittent plumes whose dynamics redistribute energy from incoherent fluctuations. It requires a modified Gradient Diffusion closure that accounts for this plume coherence to accurately model the transport in Model C.

Jocelyn: It's a lot of detail, but the main point is that non-equilibrium transport isn't just about local gradients; it depends on the specific dynamics of these plumes and their intermittency.

Subrahmanyan: The cubic scaling of flux velocity and the broad downflow wings in Model C are those key statistical signatures that we should look for when analyzing observational data from stellar convection zones.

Vera: We'll keep an eye out for how much these plume dynamics affect the energy budget in stars across different cooling regimes.

Jocelyn: We’ve got a lot of interesting stuff to dig into next on how these structures evolve over time and space.

The paper's summary: Vera: So, we're looking at how heat moves in stars using three dimensional simulations of two different convection models, Model C and Model S. This paper dives deep into those transport properties and shows a huge difference between the two regimes.

Jocelyn: Right, so the main takeaway is that when you have a cooling-driven model, like Model C, you get these intermittent plumes near the surface that create a turbulent energy flux nearly ten times higher than what we see in Model S up in the upper convection zone.

Subrahmanyan: That difference is massive. It suggests that surface cooling isn't just a minor effect; it drives a whole different kind of energy transport compared to those deeper, more stationary eddies found in Model S.

Vera: Exactly. And it’s not just the amount of heat you move; the way you move it is fundamentally different because those plumes create these broad, non Gaussian velocity distributions that we see in Model C.

Jocelyn: That's where things get interesting for observation, Vera. The paper shows that frequent high speed downflows are what dominate the transport mechanism in Model C statistically.

Subrahmanyan: And they found a specific mathematical relationship there too—that the turbulent energy flux is proportional to the cube of delta u z in that downflow regime. That cubic scaling is a very strong signature for us to look for in data.

Vera: It means we can’t just use simple average values anymore; we need to account for this intermittency and these specific statistical signatures when modeling stellar convection.

Jocelyn: The paper addresses the fact that old theories, like gradient diffusion, totally fail Model C because the mean entropy gradient is small even when there’s lots of energy moving.

Subrahmanyan: That failure points directly to a gap in our theoretical tools. They introduce a modified Gradient Diffusion approach that uses something called Time Space Double Averaging to fix this.

Vera: They correct the turbulent diffusivity by adding a term that depends on the average of u tilde dot grad u prime squared, and it turns out this correction is negative in the upper convection zone, which actually means incoherent energy is transported inward.

Jocelyn: So, their conclusion is that surface cooling generates these coherent plumes whose dynamics actively feed on and redistribute the energy from those smaller, more random fluctuations we usually just ignore.

Subrahmanyan: That’s a physically motivated framework connecting the small-scale plume dynamics to the large-scale mean transport, which is a big step forward for non equilibrium convection modeling.

Vera: The paper flags that they can't predict the exact size of this correction term without knowing how long you average over, which is a bit of a limitation for applying it right now.

Jocelyn: And they also mention that some model constants, like alpha, need to be calibrated against actual data because we just can't derive them from first principles yet.

Subrahmanyan: Looking ahead, the authors say future work needs to survey more parameter space and include rotation or magnetic fields to see if this plume-driven picture holds up outside of just the solar convection zone.

Vera: So, they’re pushing for tests using those non Gaussian statistics—the broad downflow tails and that cubic scaling—in helioseismic inversions or high cadence flow diagnostics.

Jocelyn: That sounds like a real roadmap for observational astronomy to test this new physics, moving beyond just the average properties of the flow.

Subrahmanyan: If we can confirm those specific non Gaussian features, it would provide concrete evidence that surface cooling generates coherent structures that redistribute energy in a way simple local gradients can't explain.

The paper's improvements: Tom: So, we're talking about how they fixed their math after showing those big differences between Model C and Model S transport properties. This segment looks at how they improved the modeling approach to make it more accurate.

Vera: Right, so the main improvement is introducing that modified Gradient Diffusion closure that incorporates plume dynamics directly into the math using Time Space Double Averaging.

Jocelyn: What does that actually do for us? It allows them to predict the turbulent energy flux in non equilibrium convection zones quantitatively, capturing those physical mechanisms like plume-mediated transport that standard models completely miss.

Subrahmanyan: That’s significant because it moves the modeling from a descriptive tool to a predictive one. The approach uses terms proportional to the average of u tilde dot grad u prime squared to correct the turbulent diffusivity, and this correction term is negative in the upper convection zone.

Vera: That negative term is key; it implies that incoherent kinetic energy actually gets transported inward, which enhances the effective turbulent energy transport coefficient in a way that simple models couldn't capture.

Jocelyn: So, they’ve successfully built a physically transparent framework linking those coherent plume dynamics to the mean-field transport we need to see in simulations like Model C.

Subrahmanyan: Yes, and they also trained their AI system to recognize those non Gaussian statistics in the turbulent energy flux, specifically identifying that cubic scaling relation where delta ei delta uz is proportional to the cube of delta uz in the downflow regime.

Vera: That moves us beyond just seeing broad tails; it means we can use AI to look for that specific cubic scaling relationship when analyzing observational data from stellar convection zones.

Jocelyn: And they showed that using the Time Space Double Averaging method, they can extract a coherent fluctuation variable, which they call u tilde, and this variable peaks right where the flux is largest in Model C.

Subrahmanyan: That u tilde variable serves as a really powerful diagnostic tool for quantifying exactly how much of the transport we are seeing is actually due to those coherent plumes versus just random fluctuations.

Vera: The main limitation they point out is that the magnitude of that non equilibrium correction term depends entirely on the time averaging window T you choose to define coherence, which means it’s not a single fixed number right now.

Jocelyn: And they also note that model constants like alpha still can't be predicted from fundamental equations; they have to be calibrated against actual data because we just don't have a formula for them yet.

Subrahmanyan: So, the paper gives us a better way to model non equilibrium convection, but it reminds us that we still need those calibration points and more complex physics like rotation or magnetic fields to see how this applies elsewhere.

Vera: It’s a lot of work, but the whole point is that this modified framework successfully bridges the gap between simulation results and theoretical modeling for Model C.

Jocelyn: So we have a better tool to understand why surface cooling drives energy transport differently than deep convection, even if we still need to refine the constants and look at more complex stellar environments next.

Conclusion: Tom: So, to wrap up, we’re looking at "Modeling stellar convective transport with plumes: II. Transport Properties of Locally and Non-locally driven Convection" and seeing what these results actually mean for stellar physics.

Vera: Essentially, this paper shows that surface cooling creates coherent plumes that actively redistribute energy from smaller fluctuations in a way that local gradient theories completely miss.

Jocelyn: That means the transport isn't just about a simple average of gradients; it depends on the specific dynamics and intermittency of those plumes.

Subrahmanyan: The cubic scaling relationship they found, where the turbulent energy flux is proportional to delta u z cubed in that downflow regime, provides a concrete target for observational tests.

Vera: Exactly. We can use helioseismic inversions or high cadence flow diagnostics to try and confirm that specific cubic scaling when we look at stars.

Jocelyn: If we can measure those statistics, it would give us direct evidence that the physics of surface cooling is driving these coherent structures in the way they’re modeled here.

Subrahmanyan: It confirms a non equilibrium transport process where coherent motions mediate energy flow, which is important context when we look at structure formation on larger scales.

Vera: The method they developed, combining the modified Gradient Diffusion approach with Time Space Double Averaging, gives us a way to bridge the gap between complex numerical simulations and theoretical modeling for this specific regime.

Jocelyn: It’s a great example of how incorporating non local effects into a closure can dramatically change the results you get from your simulation.

Subrahmanyan: That’s the goal; to have physical models that are robust enough to handle these non equilibrium convection zones without needing so many arbitrary parameters like alpha.

Vera: The caveat is that they still need to calibrate those constants against real data, and the correction depends on the time averaging window T you use, which limits how precisely we can predict it right now.

Jocelyn: So, while the results are powerful for predicting transport behavior in Model C, we’re still waiting for more work to tie those model parameters down with better observational constraints.

Subrahmanyan: This paper sets a clear direction for future research: surveying more parameter space and including rotation or magnetic fields to see if this plume-driven picture holds up beyond just the solar convection zone.

Vera: So, looking at "Modeling stellar convective transport with plumes: II. Transport Properties of Locally and Non-locally driven Convection," we’ve seen how surface cooling generates coherent plumes that redistribute energy in a way local gradient theories can't explain.

Jocelyn: It’s a really interesting piece because it gives us specific statistical signatures, like the cubic scaling, that we can hunt for with future observational tools.

Subrahmanyan: Confirming that cubic flux velocity scaling would be a strong confirmation of this physical picture in stellar interiors.

Department of Applied Physics, Faculty of Science, Fukuoka University · Division of Science, National Astronomical Observatory of Japan · Institute of Industrial Science, University of Tokyo

astro-ph.SR, physics.flu-dyn

Submitted: 2025-11-16

Updated: 2026-11-14

Comments: 14 pages, 17 figures, accepted for publication in MNRAS

DOI: 10.1093/mnras/stag1988

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Importance score: 90/100

The gist: Modeling stellar convective transport with plumes: II.

Key concepts

Cooling-driven model (Model C)
This convection regime is dominated by intermittent, plume-like downflows spontaneously generated near the surface due to cooling. Unlike other models, large-scale convective flows from deeper layers are absent in this scenario. These plumes are crucial for understanding the transport properties in this specific stellar environment.
Entropy-gradient–driven model (Model S)
This convection regime is characterized by large, relatively stationary eddies excited at depth. It serves as a contrasting case to Model C, representing a different physical mechanism driving convective motion. The study uses it as a baseline to highlight the distinct transport characteristics of the cooling-driven regime.
Gradient–Diffusion (GD) hypothesis
This is a conventional theoretical approach used to model turbulent transport by assuming that energy moves primarily based on local differences in entropy gradients. The paper demonstrates this simple model fails for Model C because it underestimates energy transport, proving that coherent plume dynamics are necessary for accurate description.
Time–Space Double Averaging (TSDA)
This is a statistical method used to analyze simulation data to extract the coherent component of turbulent motion, denoted as u_tilde. The paper shows that this variable peaks where the turbulent energy flux is largest in Model C, confirming that coherent plume dynamics mediate the overall transport.

Terminology

Summary

Modeling stellar convective transport with plumes: II. p. 0 Transport Properties of Locally and Non-locally driven Convection performs three-dimensional hydrodynamic simulations of two idealized regimes of stellar convection: a cooling-driven model (Model C) and an entropy-gradient–driven model (Model S). p. 0 The two regimes exhibit striking contrasts: while Model S develops large, relatively stationary eddies excited at depth, Model C is dominated near the surface by intermittent plume-like downflows that produce broad non-Gaussian velocity distributions and a turbulent energy flux that exceeds Model S by nearly an order of magnitude in the upper convection zone

Contrasting Convection Regimes

The study investigates two contrasting convection models: the cooling-driven model (Model C) and the entropy-gradient–driven model (Model S) Model S develops large, relatively stationary eddies excited at depth, while Model C is dominated near the surface by intermittent plume-like downflows that produce broad non-Gaussian velocity distributions and a turbulent energy flux that exceeds Model S by nearly an order of magnitude in the upper convection zone The two models are similar to those studied by Cossette & Rast (2016, CR16), except that CR16 included the stably stratified layer beneath the CZ, whereas we restrict the computational domain to the CZ itself In Model C, plume-like downflows are spontaneously generated near the surface, whereas large-scale convective flows from deeper layers are absent

Statistical Signatures of Plume Dynamics

The statistical properties reveal significant differences between the two regimes For Model C, the probability density of vertical velocity reveals a clear model dependence where the downflow wing is substantially broader than in Model S The joint probability distributions of vertical velocity and vertical turbulent energy flux demonstrate that frequent, high-speed downflows dominate the transport, exhibiting a nonlinear scaling Specifically, the turbulent energy flux is approximately proportional to δuz3 in the downflow regime

Failure of Local Gradient-Diffusion Models

The conventional gradient–diffusion (GD) hypothesis fails to describe Model C accurately The GD model underestimates transport in the upper CZ because the mean entropy gradient is small, even though vigorous plumes carry substantial energy This discrepancy clearly demonstrates that the cooling-driven convection cannot be described by a simple gradient-diffusion closure

Development of a Modified Closure

To address this failure, the authors introduce a modified Gradient–Diffusion approach incorporating plume dynamics via the Time–Space Double Averaging (TSDA) method They propose a modified GD closure where the turbulent diffusivity is corrected by a plume-mediated term proportional to ⟨(u˜·∇)u′2⟩h This correction term is negative in the upper CZ, implying an inward transport of incoherent kinetic energy that enhances the effective turbulent energy transport coefficient

Astrophysical Implications

The results suggest that surface cooling generates coherent, intermittent plumes whose dynamics feed on—and redistribute—the energy of incoherent fluctuations The predicted non-Gaussian statistics, including broad downflow tails and cubic flux–velocity scaling, provide concrete targets for observational tests using helioseismic inversions and high-cadence flow diagnostics This modified closure offers a physically motivated extension applicable to non-equilibrium stellar convection zones

Limitations and Future Directions

Two main limitations are noted: the magnitude of the non-equilibrium correction depends on the time-averaging window T used to define coherence Furthermore, model constants like α cannot yet be predicted from fundamental equations and must be calibrated against data Future work will survey parameter space and include rotation/MHD to clarify applicability beyond the solar CZ

Summary of Key Findings

The main findings include:

(i) Contrasting convection morphologies:

Model C is dominated in the upper CZ by intermittent, plume-like downflows generated by surface cooling, whereas Model S exhibits larger, more stationary eddies excited at depth

(ii) Marked non-Gaussianity in Model C:

Probability densities of the vertical velocity develop broad downflow wings in Model C, and the joint probability distributions of δuz and δeiδuz demonstrate that frequent, high-speed downflows dominate the transport

(iii) Enhanced turbulent energy flux near the surface:

The vertical turbulent energy flux in Model C exceeds that of Model S by up to an order of magnitude in the upper CZ

(iv) Breakdown of local GD closures:

While the GD approximation reproduces Model S fairly well, it underestimates the flux in Model C by nearly an order of magnitude in the upper CZ

(v) Role of coherent motions and TSDA:

Using TSDA, the coherent component u˜z peaks where the flux is largest in Model C, demonstrating that plume coherence mediates transport

(vi) A modified non-equilibrium closure:

A modified GD closure successfully reproduces both the amplitude and depth-dependence of the flux in Model C when parameters are appropriately specified

Conclusion

The TSDA analysis together with the success of the modified closure points to a consistent physical picture: surface cooling generates coherent, intermittent plumes whose dynamics feed on—and redistribute—the energy of incoherent fluctuations The resulting non-equilibrium transport cannot be captured by a purely local, gradient–based flux; instead, it depends on plume coherence and intermittency By embedding plume dynamics into the closure, the modified GD framework successfully bridges the gap between numerical simulations and theoretical modeling

Final Caveats

The present closure still depends on the choice of T and on a calibrated α Developing strategies to tie these parameters to observable plume statistics will be an important direction for future work

Acknowledgements

Numerical computations were carried out on Cray XC50 at Center for Computational Astrophysics (CfCA), National Astronomical Observatory of Japan (NAOJ) This work was supported by the Research Institute of Stellar Explosive Phenomena at Fukuoka University & also from the University grant No.GR2302, and also by JSPS KAKENHI Grant Number (JP17H06364, JP18H01212, JP18K03700, JP21K03612, JP21H01088, JP23K22494, JP23K25895, JP23K03400, JP24K00631 and JP25K07374) This research was also supported by MEXT as “Program for Promoting researches on the Supercomputer Fugaku” (Structure and Evolution of the Universe Unraveled by Fusion of Simulation and AI; Grant Number JPMXP1020230406) and JICFuS The National Institutes of Natural Sciences (NINS) program for crossdisciplinary study on Turbulence, Transport, and Heating Dynamics in Laboratory and Solar/Astrophysical Plasmas: “SoLaBo-X” supports this work

Data Availability

The data underlying this article will be shared on reasonable request to the corresponding author The paper is typeset from a TEX/LATEX file prepared by the author The references list includes authors such as Masada Y., Sano T., and Yokoi N. p. 1 The work is supported by grants from Fukuoka University and the National Astronomical Observatory of Japan The paper was accepted under XXX and received in original form ZZZ This paper has been typeset from a TEX/LATEX file prepared by the author The authors are Youhei Masada, Tomoya Takiwaki, and Nobumitsu Yokoi The paper was accepted under XXX. p. 0 Received YYY; in original form ZZZ This paper is typeset from a TEX/LATEX file prepared by the author The authors are Youhei Masada, Tomoya Takiwaki, and Nobumitsu Yokoi The paper was accepted under XXX. p. 0 Received YYY; in original form ZZZ

References

The references list includes Ahlers G., Grossmann S. p. 9, Lohse D., 2009, Reviews of Modern Physics, 81, 503 Anders E. p. 9 H., Brown B. P., 2017, Physical Review Fluids, 2, 083501 Birch A. p. 9 C., Proxauf B. p. 1, Duvall T. p. 9 L., Gizon L., Hanasoge S. p. 1, Hindman B. p. 9 W., Sreenivasan K. p. 9 R., 2024, Physics of Fluids, 36, 117136 Böhm-Vitense E. p. 9, 1958, Z. Astrophys., 46, 108 Boussinesq J. p. 9 V., 1877, Essai sur la théorie des eaux courantes. p. 9 Paris: Impr. p. 9 nationale Bradshaw P. p. 9, 1974, Nature, 249, 135 Brandenburg A. p. 9, 2016, ApJ, 832, 6 Brummell N. p. 9 H., Clune T. L., Toomre J. p. 1, 2002, ApJ, 570, 825 Bushby P. p. 9 J., Käpylä P. p. 1 J., Masada Y. p. 0 p. 0, Brandenburg A. p. 1, Favier B. p. 9, Guervilly C. p. 9, Käpylä M. p. 1 J., 2018, A&A, 612, A97 Chandrasekhar S. p. 9, 1961, Hydrodynamic and hydromagnetic stability Christensen-Dalsgaard J. p. 9, et al., 1996, Science, 272, 1286 Cossette J. p. 9-F., Rast M. P., 2016, ApJ, 829, L17 Rieutord & Zahn (1995), A&A, 296, 127 Rincon F. p. 9, Rieutord M. p. 1, 2018, Living Reviews in Solar Physics, 15, 6 Sano T. p. 9, Inutsuka S., Miyama S. M., 1999, in Miyama S. p. 2 M., Tomisaka K., Hanawa T., eds, Astrophysics and Space Science Library Vol. p. 0 240, Numerical Astrophysics. p. 9 p. 383 Schrijver C. p. 1 J., Siscoe G. p. 1 L., 2013, Heliophysics 3 Volume Paperback Set Spiegel E. p. 1 A., 1971, ARA&A, 9, 323 Spruit H. p. 3 C., Nordlund A. p. 3, Title A. M., 1990, ARA&A, 28, 263 Stefan J. p. 1 T., Kosovichev A. G., Guerrero G., Stejko A. M., 2025, arXiv e-prints Stein R. p. 0 F., Nordlund Å. p. 3, 1998, ApJ, 499, 914 Stix M. p. 0, 2002, The sun: an introduction Stix M. p. 0, 2004, The Sun: An Introduction Yokoi N. p. 0, 2018, Journal of Plasma Physics, 84, 775840603 Yokoi N. p. 1, Masada Y., Takiwaki T. p. 0, 2022, MNRAS, 516, 2718 The paper was accepted under XXX. p. 0 Received YYY; in original form ZZZ

Key Words

The key words are convection – hydrodynamics – turbulence – stars: interiors – Sun: interior This paper is typeset from a TEX/LATEX file prepared by the author The authors are Youhei Masada, Tomoya Takiwaki, and Nobumitsu Yokoi The paper was accepted under XXX.

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Improved AI systems can now quantitatively predict turbulent energy flux in non-equilibrium convection zones by using a modified Gradient-Diffusion closure that incorporates a non-equilibrium turbulent diffusivity term, as shown in equation 17. This allows the system to capture physical mechanisms like plume-mediated transport that conventional models miss.

  2. The improved AI can perform time and space double averaging (TSDA) on simulation data to extract the coherent fluctuation variable u tilde, which peaks where the flux is largest. This diagnostic variable serves as a key tool for quantifying plume contribution to transport in cooling-driven regimes.

  3. AI systems can now implement a modified Gradient-Diffusion approach where turbulent diffusivity is corrected by terms proportional to ⟨(u tilde · grad) u prime 2⟩ h, achieving quantitative agreement with simulation results in Model C. This provides a physically transparent framework linking coherent plume dynamics to mean-field transport.

  4. The AI can be trained to recognize non-Gaussian statistics in turbulent energy flux, specifically identifying the cubic scaling relation ⟨delta ei delta uz⟩ proportional to delta uz cubed in the downflow regime of Model C. This capability moves beyond simple Gaussian modeling for stellar convection.

Sources

Related papers